A patch algorithm for fast registration of distortions

This note describes an algorithm for registration of non-linear distortions in two-dimensional (2D) images. The algorithm solves an integrated form of linearized image matching equation over a set of patches in the image domain. The distortion field is represented as a linear combination of multinomials in the image-plane coordinates. Prior to registration, the images are convolved with band-pass filters that are implemented at a coarse-to-fine hierarchy of spatial scales (specified at input by the user); and some heuristics are given for choosing these scales. Images registered at a coarse scale are used as input approximations for deriving images registered at a subsequent finer scale of resolution. With this prescribed hierarchy, the algorithm is non-iterative; it is also robust to noise, and can in principle accommodate arbitrary local or global distortions. The results indicate that non-linear registration does improve the match between images.

[1]  J C Mazziotta,et al.  Automated image registration: II. Intersubject validation of linear and nonlinear models. , 1998, Journal of computer assisted tomography.

[2]  Daniel Rueckert,et al.  Nonrigid registration using free-form deformations: application to breast MR images , 1999, IEEE Transactions on Medical Imaging.

[3]  D. Hill,et al.  Medical image registration , 2001, Physics in medicine and biology.

[4]  Berthold K. P. Horn,et al.  Determining Optical Flow , 1981, Other Conferences.

[5]  Laveen N. Kanal,et al.  Gradient Based Image Motion Estimation Without Computing Gradients , 1997, International Journal of Computer Vision.

[6]  J. Hajnal,et al.  A Registration and Interpolation Procedure for Subvoxel Matching of Serially Acquired MR Images , 1995, Journal of computer assisted tomography.

[7]  Karl J. Friston,et al.  Spatial registration and normalization of images , 1995 .

[8]  Margaret E. Daube-Witherspoon,et al.  A head motion measurement system suitable for emission computed tomography , 1997, IEEE Transactions on Medical Imaging.